Re: gnome-vfs code freeze break request - http method crasher

["Murray Cumming" <murrayc murrayc com>]

> On Tue, 2004-09-07 at 10:10 +0200, Murray Cumming wrote:
>> > Hi!
>> >
>> > I fear I have to ask for another code freeze break for the http neon
>> > method. I wouldn't do it but the patch fixes a really bad bug in the
>> > http neon method. The problem with the current code is that it assumes
>> > that ne_read_response_block returns only positiv values or 0. But in
>> > fact it can return -1 on a IO error. In exactly that case calling
>> > ne_end_request () will crash the method. In other words one must not
>> > call ne_end_request if ne_read_response_block returns -1. So the patch
>> > does correct two important things. 1) It changes "n" from
>> > GnomeVFSFileSize to ssize_t so n can be negative. 2) It catches the
>> case
>> > that n == -1 and avoids calling ne_end_request in that case but sets
>> the
>> > results to GNOME_VFS_ERROR_IO (what is exactly what happened then).
>> > I am sorry for the complicated patch but I think this is the cleanest
>> > way to fix it. If you have any other questions feel free to ask.
>> > It also removes the check for bytes_read being NULL because gnome-vfs
>> > will ensure that it inst for us.
>>
>> Thanks, but we need to know what bug the user currently sees. When will
>> the user experience this bug, and what will it look like?
>
> If you use a proxy, it will segfault when you read a file.

Aha, it doesn't get much worse than that. 1 of 2 approvals, and sorry for
not noticing the patch.

Murray Cumming
murrayc murrayc com
www.murrayc.com
www.openismus.com